Repeatability
High
The task is a one-time script generation with a fixed structure: parse XML, map fields, validate, log errors. The logic is deterministic and the same pattern applies across all three suppliers once schemas are provided.
Ambiguity Tolerance
High
Success criteria are concrete: normalized JSON output, required-field validation, error logs with line numbers, and throughput at 5K+ orders/day. A generated script can be tested against these criteria objectively.
Data & Tool Availability
Medium
The agent needs the actual XML schema samples or example feeds from all three suppliers to generate accurate field mappings. Without those, it can produce a solid template but field-level accuracy will require human verification.
Error Cost
Medium
A bug in normalization logic could silently mismap order fields (e.g., quantity vs. unit price), causing downstream fulfillment errors. However, the error-logging requirement and the fact that this is a script to be reviewed before deployment significantly reduce blast radius.
Human Judgment Required
Low
Field mapping decisions are largely mechanical once schemas are known. The only judgment calls are around ambiguous or missing fields, which the error-logging requirement explicitly routes to human review anyway.